Klarna Cut 40% of Its Workforce for AI. Now It's Rehiring. Here's What That Tells Us.

The Klarna Experiment

In 2024, Klarna CEO Sebastian Siemiatkowski was one of the most confident voices in the AI-replaces-humans conversation. The Swedish digital bank and payments company, with a global customer base and a reputation for tech-forward thinking, was positioned as the perfect test case. Siemiatkowski was direct: AI could take over the jobs. The company froze all hiring and began cutting aggressively.

The workforce dropped from 5,500 to 3,400 — a reduction of approximately 40%. The stated replacement was an AI chatbot deployed to handle customer service at scale, performing, the company claimed, the work of more than 700 agents. This was supposed to be the proof of concept. The moment that would validate the entire theory: that AI could absorb the operational core of a modern financial services company and do it cheaper, faster, and without the friction of managing people.

It was, for a time, the story the AI industry wanted told. Klarna became a reference point in presentations and board meetings. If a payments company could cut 40% of its staff and keep running, the logic went, anyone could.

What the Chatbot Actually Did

The chatbot worked. That part is worth acknowledging, because the failure of the Klarna experiment is not that AI is useless in customer service. For predictable, repetitive, low-stakes queries — password resets, balance inquiries, standard payment confirmations — the system performed. Volume was handled. Response times were fast. The economics looked good on the specific slice of interactions the chatbot was designed for.

The problem was everything outside that slice. Complex disputes. Fraud investigations. Account anomalies that required judgment, context, and the ability to hold two pieces of contradictory information simultaneously and decide between them. The chatbot could not do this. Not reliably. Not in the ways that customers needed when something had gone genuinely wrong with their money.

Customers grew frustrated. Trust in the company declined. The friction that Klarna had eliminated by removing human agents was replaced by a different kind of friction — the cold, looping dead-end of an AI that cannot understand what you actually need. Engineers and marketing staff were pulled to answer customer calls. The spectacle of data scientists triaging payment disputes is not the efficiency story anyone wanted to tell. But it is the one that played out.

The Reversal

Klarna reversed course. The rehiring began. The quiet acknowledgment that the experiment had not delivered what was promised was more telling than any public statement — because Klarna was not alone, and the company's leadership knew it.

Across the tech sector, companies that had made aggressive AI-for-headcount bets were doing the same arithmetic and arriving at the same conclusion. The cost of reversing the layoff — recruiting, retraining, rebuilding institutional knowledge that had walked out the door — was substantial. The lesson that should have been obvious in advance was becoming obvious in practice: when you fire people because you think a machine can replace them, and the machine cannot, you do not get those people back easily or cheaply.

The pattern is consistent enough that it now has its own category in analyst forecasting. This is not a collection of isolated mistakes. It is a systemic misreading of what AI can do at the edges of complex human interaction — and the correction is already underway.

What Forrester Sees Coming

Forrester's research into AI adoption patterns produced a prediction that is striking in its specificity: 50% of all AI-related layoffs will be reversed by 2027. The firm also found that 55% of employers already regret the decision to cut staff in the name of AI efficiency — a number that suggests the reversal is not some distant possibility but a process already in motion at the level of individual regret, before the formal rehiring has caught up.

The underlying logic is straightforward and painful. Companies that eliminated staff to invest in AI acceleration fired the people who understood their systems, their customers, and the edge cases that no model is trained to handle. They eliminated institutional knowledge in exchange for speed on the median query. And when the tail of the distribution — the hard problems, the unusual situations, the moments that actually determine customer loyalty — needed addressing, there was no one left who knew how.

MIT's parallel finding reinforces this picture: of 300 public AI implementations, only 5% showed significant profit impact. The other 95% consumed resources without producing measurable return. The companies in that 95% are not going to keep funding the experiment indefinitely. When the pivot comes, the demand for human expertise will be real — and it will be urgent.

The AI Reversal Scoreboard

What This Means for Workers

The people who were let go are about to negotiate from a position of strength — and the companies that fired them largely do not understand this yet. The workers who built the infrastructure, managed the edge cases, held the institutional knowledge, and understood the customer in all their complexity are not going to be replaced by the next version of the model. They are going to be recruited back. And this time, they will know what their absence cost.

The skills that will command the highest premium in that rehiring wave are the ones that sit at the boundary between human judgment and AI capability: AI prompt engineering, workflow automation design, AI auditing and quality assurance, and the ability to identify precisely where an AI system breaks down and why. These are not exotic credentials. They are learnable, and the two-year window before Forrester's predicted reversal is exactly the right amount of time to build them.

The Klarna story is not a cautionary tale about technology. It is a cautionary tale about impatience and the temptation to mistake a tool for a replacement. The people who spent the layoff period building hybrid human-AI skills will return to the workforce not as the people who were fired, but as the people who know how to make both things work together. That is a different and considerably more valuable position to negotiate from.